Computing cavities, channels, pores and pockets in proteins from non-spherical ligands models

Motivation: Identifying protein cavities, channels and pockets accessible to ligands is a major step to predict potential protein-ligands complexes. It is also essential for preparation of protein-ligand docking experiments in the context of enzymatic activity mechanism and structure-based drug design. Results: We introduce a new method, implemented in a program named CCCPP, which computes the void parts of the proteins, i.e. cavities, channels and pockets. The present approach is a variant of the alpha shapes method, with the advantage of taking into account the size and the shape of the ligand. We show that the widely used spherical model of ligands is most of the time inadequate and that cylindrical shapes are more realistic. The analysis of the void parts of the protein is done via a network of channels depending on the ligand. The performance of CCCPP is tested with known substrates of cytochromes P450 (CYP) 1A2 and 3A4 involved in xenobiotics metabolism. The test results indicate that CCCPP is able to find pathways to the buried heminic P450 active site even for high molecular weight CYP 3A4 substrates such as two ketoconazoles together, an experimentally observed situation.